10 Pitfalls to Avoid When Designing Any Additional Liquidity Requirements

The bank failures in spring 2023 have increased recognition of the importance of the discount window for liquidity risk management. The focus on discount window preparedness is a welcome development, but any new liquidity requirement should be developed thoughtfully and with input from all stakeholders. This note describes 10 pitfalls the banking agencies should seek to avoid when making changes to liquidity regulations.

There appear to be several different approaches floating around for encouraging banks to be more ready to borrow. A month ago, a report by the Group of Thirty recommended that each bank be required to have discount window borrowing capacity that, when combined with deposits at the Fed, exceed uninsured deposits and short-term funding. A week later, Acting OCC Comptroller Mike Hsu recommended that banks be required to have discount window borrowing capacity and reserves available in an amount at least as large as potential five-day outflows under severe liquidity stress. Another possibility that frequently arises in conversations is requiring each bank’s discount window borrowing capacity and deposits at the Fed to exceed 40 percent of the bank’s uninsured deposits. A bit further back, two weeks before SVB failed, BPI recommended that the banking agencies put increased emphasis on discount window borrowing capacity when evaluating the liquidity condition of commercial banks through the examination process (in particular, the internal liquidity stress tests) and conduct a holistic review of liquidity regulations. 

The common thread across these and other related proposals is a recognition that a bank that is prepared to borrow from the discount window is more liquid than one that isn’t, and that weaving that notion into regulatory and supervisory assessments of bank liquidity would make those assessments more accurate and increase incentives for banks to be prepared. Moreover, being allowed to resolve greater needs for short-term contingency funds with discount window capacity rather than reserve balances allows banks to devote more of their balance sheet to lending to businesses and households – loans that can then be pledged to the Federal Reserve as collateral – rather than expanding even further their loans to the government in the form of even larger deposits at the Fed. 

For example, in a speech on February 12, Andrew Bailey, Governor of the Bank of England, rejected the idea of responding to the rapid deposit outflows observed in the spring by simply requiring banks to hold ever larger amounts of government debt and deposits at the central bank:

I don’t think the answer is that simple. This would amount to saying that banks should self-insure more. But it would move banking significantly towards a narrow bank model which would disrupt the process of credit creation – lending – in the economy, with negative economic consequences. Supporting economic activity is an important part of so-called fractional reserve banking, in which only part of deposit liabilities are backed by banks in highly liquid assets. Changing this arrangement does not strike me as the right way to go. The alternative, and better way I think, is to supplement the existing liquidity regime with more ready access for banks to liquidity insurance at the Central Bank, which is appropriately priced and risk managed. This could be alongside more targeted adjustments to the liquidity regime, perhaps aimed at firms with more vulnerable business models, and for banks to be prepared to access liquidity insurance to monetise assets at speed.

Enabling banks to shift away from deposits at the Fed as a significant component of their liquid assets would have the added benefit of allowing the Fed to get smaller and less involved in the financial system on a day-to-day basis.  Indeed, if the Fed were to return to its pre-GFC approach to conducting monetary policy, interbank markets, which have been decimated by the deluge of liquidity caused by the Fed’s inflated balance sheet, would recover. Claudio Borio, head of the Monetary and Economic Department at the Bank for International Settlements and a global expert on central bank operating frameworks, recognized the value of returning to such a situation in a recent Macro Musings podcast:

Would you like to have a system in which the central bank is a backstop, or would you like to have a system in which the central bank is the mass market maker of first resort, so last resort versus first resort?… I think that having a system in which the central bank is a backstop, and a system in which the first line of defense against demands on liquidity is an interbank market, that to me sounds [like], on balance, a better system.

Nevertheless, there are some pitfalls to avoid when designing such a system.

1. Be Clear on the Objective

The tragedy of last spring was not that Silicon Valley and Signature Banks failed – banks are supposed to fail occasionally. The tragedy was that the failures resulted in a sufficiently a high risk of contagion and a broader banking panic that the government decided to invoke the systemic risk exception to least cost resolution and guarantee the uninsured depositors of the failed bank and promise to bail out the uninsured depositors of any subsequent banks that failed. The Fed also opened the emergency Bank Term Funding Program in which the Fed provided undercollateralized loans at a below-market rate, exactly the opposite of the traditional lender of last resort guidelines to lend against abundant collateral at an above-market rate.

The run on Silicon Valley Bank was what is called a rational run. Because of the losses the bank had experienced on its government securities portfolio, it did not have the resources to meet a run by its depositors in the best of circumstances, even if all its assets had been pledged to the discount window. Nevertheless, it is undoubtedly true that if SVB had been prepared to use the window, its failure would have been much more orderly, and perhaps would not have triggered an immediate run and subsequent disorderly failure of Signature Bank and broader concerns about other similar banks. No bank can survive a complete loss of investor confidence, especially one with embedded losses that have wiped out its capital. Liquidity regulations have not been designed to ensure a bank can weather such a storm. One possible objective of the contemplated new discount window requirements is achieving orderly resolution rather than limiting liquidity risk.

An alternative potential objective is reducing the likelihood of contagion to sound banks. What seemed to drive the need for the extraordinary government action was not an unwillingness to let SVB’s creditors take their lumps as much as it was a judgment that many other sound banks also were vulnerable to runs and did not have sufficient discount window collateral to meet a depositor run if one were to materialize. If the objective is to avoid such circumstances in the future, rather than calibrating any new standard to SVB, it should instead be calibrated to the level sufficient to enable sound banks to survive a run. Put another way, what amount of preparedness by sound banks would have given the leadership of the banking agencies confidence that SVB could have been allowed to fail without having to bail out the stablecoin issuer and Silicon Valley investors and businesses that were its uninsured depositors?

2. Do Not Repeat the Mistakes of the Past

Some of the new requirements being considered are nearly indistinguishable from reserve requirements. Reserve requirements require banks to maintain reserves – vault cash and deposits at the Fed – in an amount at least as large as a specified fraction of deposits. Adding discount window borrowing capacity to the list of allowable sources of funding does not change the fundamental nature of the requirement. It’s been understood for well over a century, however, that arithmetic and rigid reserve requirements reduce rather than enhance the resilience of the banking system. The problem is that if a bank is just meeting its requirement and uses some of its reserves to meet a deposit outflow, it immediately falls out of compliance with the requirement. As we discussed in “The Middle Course: What Fed History Teaches Us About Liquidity Requirements,” such requirements contributed to the Banking Panic of 1907 by leading banks to shut their doors well before their reserves were exhausted.

Relatedly, there is widespread recognition that banks are unwilling to use the high-quality liquid assets they are required to maintain to meet the liquidity coverage ratio to meet a liquidity need lest they fall out of compliance with the regulation. The Basel Committee published two reports following the onset of the COVID-19 pandemic showing bank reluctance to use their stockpile of HQLA. The Bank of England and Prudential Regulation Authority also issued a discussion paper, the responses to which indicated “most respondents agreed…that banks are reluctant to draw on their stock of HQLA in periods of unusual liquidity pressure.” Similarly, a survey of BPI treasurers found that none would be willing to use their HQLA if it meant falling out of compliance, and some reported pulling back from providing liquidity to others and selling assets into illiquid markets to avoid violating their HQLA, exactly the outcome the LCR was intended to avoid. While allowing banks to meet heightened expectations with discount window borrowing capacity rather than even more unusable HQLA is sensible, the requirements will only enhance the resilience of the banking system if the resources can and will be used.

One way to avoid the reserve requirement/usability problem is to build any new requirements into expectations for contingency plans rather than establishing yet another ratio that needs to be satisfied. For example, each bank could be required to include a new, one-week horizon in its internal liquidity stress test that assumes the bank experiences severe funding stress but is allowed to plan on meeting that outflow by borrowing from the discount window. No one could fault a bank for using its contingency plan when it is in a contingency. 

3. Please, No More Ratios

Ratios just cause unnecessary trouble, and it is unsound to divide by a flow (see Cetina and Gleason (2015)). That is, if the requirement is that discount window borrowing capacity and reserve balances exceed projected short-term outflows under severe stress, don’t express it as a requirement that the funds divided by the outflows is greater than 1. If it is absolutely necessary to design the new requirement as a ratio, consider adopting the “dynamic LCR” form described in “A Modest Change to the LCR That Could Substantially Improve Financial Stability.” Specifically, when a bank experiences a run, reduce its liquidity requirement dollar-for-dollar by the amount of the run. Doing so prevents a bank from falling out of compliance by using its resources to address the run.

4. Repos Provide Their Own Collateral if They Don’t Roll

Repo borrowing against Treasuries, Agencies, or Agency MBS should either not be included in the measure of potential short-term funding need or the repoed security should be added to discount window borrowing capacity, after applying the appropriate haircut. If the repo borrowing does not roll over, the bank gets the security back and can pledge it to the discount window that same day. This treatment would be consistent with the treatment of repos in the liquidity coverage ratio. 

5. Reduce the Stigma Associated With Using the Discount Window

Another related problem is that even if banks were generally willing to use their contingent sources of liquidity, they remain specifically unwilling to borrow from the discount window because of the severe stigma associated with doing so. Thus, a necessary complement to requiring banks to have a certain amount of discount window borrowing capacity is reducing the stigma associated with borrowing. While it might be tempting to think it doesn’t matter because a bank will borrow if it must, a key objective of liquidity requirements is leaving banks sufficiently confident in their liquidity situation that they will not pull back from lending to others or sell assets at firesale prices, actions that can turn liquidity strains into a broader crisis. 

There has a been a stigma associated with using the discount window since the 1920s, and it exists in part because the Fed has often taken steps to boost it (“Discount Window Stigma: We Have Met the Enemy, and He Is Us”). Stigma increased astronomically after the GFC when borrowing was equated with having received a bailout. While there is no easy way to reduce discount window stigma, there are some minimum necessary steps. Taking a page from the Bank of England’s playbook, the leadership of the Fed and other banking agencies need to educate the public, Congress, bank examiners and bank investors that borrowing from the discount window is a business decision of the borrowing bank and neither a bailout nor an indication that the bank is in trouble (see here). The Fed could follow the BoE by conducting a liquidity stress exercise in which banks and bank examiners jointly practice making use of the discount window (see here) and by assuring banks that they are not expected to first rundown their HQLA before borrowing from the window (see here). The Fed could also imitate the BoE’s QT plan and shrink until discount window borrowing picks up. 

Moreover, as discussed further in the eighth pitfall, the banking agencies could allow banks to recognize discount window borrowing capacity in internal liquidity stress tests and resolution liquidity requirements. Efforts by the banking agencies to convince banks that they should be willing to borrow from the discount window ring hollow if banks are told that contemplating such borrowing is verboten when the bank tests its liquidity needs under stress.

6. Don’t Use Liquidity Requirements to Fix Other Problems

Banks can book securities in three categories: trading, investment account available-for-sale and investment account held-to-maturity. Trading account securities are recorded at fair/market value and gains and losses on the securities are included in income. AFS securities are recorded at fair value. Unrealized gains and losses on AFS securities do not pass through income but do increase or decrease equity. Equity for regulatory capital purposes is unaffected except for a very large bank. HTM securities are booked at par value (essentially purchase price) and unrealized gains and losses do not pass through income or influence equity. However, if a bank sells any of its HTM securities, all of its HTM securities are immediately reclassified as AFS and marked to market.

HQLA that is classified as HTM continues to count as HQLA for liquidity regulation purposes, although at market, not par, values. Even though a bank cannot sell its HTM HQLA, the securities remain highly liquid. The bank can repo the securities in a private transaction or at the Fed SRF, and the bank can use the securities to collateralize a discount window loan. In reality, it would be a rare circumstance anyway that a bank would sell HQLA to raise cash quickly.

Consequently, if the banking agencies decide to take steps to limit banks’ ability to classify investment account securities as held-to-maturity, liquidity regulations are not the appropriate vehicle for doing so any more than liquidity regulations are the appropriate mechanism to limit credit risk. Departing from reality to accomplish an unrelated objective makes liquidity regulations less accurate and so less beneficial. It also opens the door to unintended consequences. 

7. Give Notice and Seek Comment Through an Advance Notice of Proposed Rulemaking

Designing a new requirement takes careful thought and thorough analysis. For example, because FHLBs secure their advances using a blanket lien on all bank assets, a requirement that banks pledge a lot more collateral to the discount window will likely have profound consequences on how FHLBs do businesses, consequences that need to be understood in advance. As another example, the depositors that ran from SVB were all in the same line of business and tightly interconnected. Rushing to link a new requirement to undifferentiated uninsured deposits could miss the actual source of risk. And, of course, it raises legal issues to implement a new rule without notice and comment, including a rule applied through systematic changes in examiner expectation. Indeed, in light of the novelty of the rule being broadly contemplated, an advance notice of public rulemaking would be the appropriate first step.

8. Don’t Create an Inferior Rule Just So It Can Be Applied Uniformly to Institutions of All Sizes

Many smaller banking institutions are also funded in large part with uninsured deposits and have invested in longer-term assets. When banking agency leadership contemplated the possibility of a broad banking panic last spring, no doubt they were concerned about the possibility of runs on smaller banks as well as larger banks. If so, then they may be seeking to design a new rule that can be applied to banks of all sizes.

However, larger banks are subject to stricter and more comprehensive liquidity requirements and provide the banking agencies with much more information either daily or monthly. It would needlessly punish larger institutions and result in an inferior outcome if those added requirements and information were not used when sizing the larger banks’ very short-term cash needs under severe stress. For example, deposits used by a business for its day-to-day operations are stickier than additional deposits that the business maintains. Larger institutions report operational and nonoperational deposits separately, but smaller institutions do not. Larger banks should not be subject to a requirement based on the undifferentiated deposits simply because smaller banks do make the differentiation. As another example, as noted, building discount window capacity requirements into contingency plans including ILSTs may be the most effective way to ensure banks will make use of the capacity, while requiring that the capacity equal a percentage of uninsured deposits will likely make the banking system less rather than more resilient. The latter, inferior approach should not be adopted just so that the exact same rule can be applied to smaller banks that do not perform ILSTs and larger banks that do.

9. Make Conforming Changes to the Rest of the Liquidity Assessment Framework

There are several aspects of the regulatory and supervisory liquidity risk management framework that would become illogical once banks are required to have discount window borrowing capacity sufficient to meet short-term cash needs under severe stress. Many of them can be adjusted while remaining compliant with the international standard for the LCR.

  • Although the international standard for the LCR is designed to ensure that banks have sufficient liquidity at the end of 30 days of stress, the U.S. version is designed to ensure that banks can make it to 30 days. In particular, the U.S. standard includes a “maturity mismatch add-on” which increases required HQLA to correspond to the peak need over 30 days. Banks should be allowed to plan on meeting those intra-30-day needs using discount window borrowing capacity in addition to HQLA. 
  • Banks should be allowed to assume in their ILSTs that they will monetize their HQLA using the discount window or the Fed’s standing repo facility in addition to sales or private repo. Large banks are required to have contingency funding plans (CFPs) and also have monetization plans for the HQLA. Although banks can include the discount window and SRF in their CFP, at least up until recently, they have not been allowed to point to the discount window or SRF as their monetization plan.  If banks are required to have discount window borrowing capacity – capacity that would include any HQLA that had been pledged to the window – it would be illogical to not allow banks to plan on monetizing their HQLA using the discount window.
  • The agencies by regulation or guidance should require examiners to drop their preference for reserve balances rather than other forms of HQLA such as Treasury securities. If the bank has the capacity to meet several days of stress outflows using discount window borrowing capacity, it also has time to convert its Treasury securities into cash through private repo, the SRF or the discount window. As then-Vice Chair for Supervision Quarles noted in the November 2018 FOMC meeting, removing such preferences would allow the Fed to get smaller than would be possible otherwise: “…if some of the anecdotes are true that we hear of supervisors informally insisting that banks meet much of their liquidity requirements with reserves rather than other forms of high-quality liquid assets (HQLA)—and I think those reports are probably exaggerated, but neither do I think that they are fabricated—we should push against that informal regulatory intensification of the demand for reserves.”[1]
  • For some GSIBs, the most binding liquidity requirements are those associated with resolution, RLAP and RLEN. RLAP estimates the liquidity needs of each material subsidiary of a GSIB BHC over a period of at least 30 days under the assumption that liquidity from some other subsidiaries has been ringfenced, for instance, by a foreign supervisor.  RLEN is an estimate of the liquidity that would be necessary to successfully execute a GSIB BHC’s resolution plan, including stabilizing each material subsidiary after the parent BHC declares bankruptcy. The Fed and FDIC have provided guidance that states a bank may only assume that it borrows from the discount window for a few days at most. That guidance should be revisited in light of any new requirements that banks be prepared to make use of the discount window when under severe funding stress.[2]

10. Eventually, Revisit the International Standard for the LCR

If banks are required to be prepared to meet short-term severe cash outflows using reserve balances and discount window capacity, then the LCR, which requires banks to meet 30-day outflows without making use of the discount window, may eventually need to be revisited.  The G30 report, for example, states that if its recommendations were adopted, “[i]t would do away with the LCR and HQLA requirements.”  While doing away with the LCR might be too radical, at a minimum its treatment of central bank borrowing needs to be rethought.


[1] Transcript of the November 2018 FOMC meeting, pp. 36-37.  https://www.federalreserve.gov/monetarypolicy/files/FOMC20181108meeting.pdf

[2] For further discussion see the BPI note “Central Bank Contingency Funding in Resolution Plans,” August 1, 2023.  https://bpi.com/central-bank-contingency-funding-in-resolution-plans/